TypoPilot // 900-example evaluation

fine-tuned Llama 3.2 1B · Q4_K_M · CPU

Every test-suite output from the final native CPU run. The center column is the target reference—not necessarily the only acceptable correction. Use the filters to inspect exact matches, remaining errors, meaning-changing overcorrections, and individual categories.

error in inputfixed correctly by the modelstill wrong, missing, or wrongly changedextra context generated beyond the correction
900 of 900 shown